基于深度学习图像描述子的三维彩色点云配准  

3D color point cloud registration based on deep learning image descriptor

作　　者：李佳骏 安毅 秦攀 顾宏[1] LI Jiajun;AN Yi;QIN Pan;GU Hong(School of Control Science and Engineering, Dalian University of Technology, Dalian 116024, China)

机构地区：[1]大连理工大学控制科学与工程学院,辽宁大连116024

出　　处：《大连理工大学学报》2021年第3期316-323,共8页Journal of Dalian University of Technology

基　　金：国家自然科学基金资助项目(81872247,61673083);山西省科技重大专项揭榜项目(20191101014).

摘　　要：不断发展的激光扫描技术使得获取三维空间中的彩色点云信息更加方便.但是,如何将多个采集点生成的彩色点云数据统一在同一个坐标系下,构建一个完整的数据模型仍是一个挑战.因此,提出了一种基于深度学习的图像描述子,将其应用于三维彩色点云配准中,能够以较高精度获取点云配准的初始位姿.首先,根据点云和图像之间的一一对应关系,将三维彩色点云投影为图像;其次,使用卷积神经网络提取关键点邻域的局部特征,结合方向梯度直方图,形成组合描述子;再次,根据计算出的组合描述子计算点云的匹配点对,得出点云间的转换关系,实现点云粗配准.以实际的三维彩色点云数据与多种配准算法进行对比,验证了所提方法的有效性.The continuous development of laser scanning technology makes it convenient to obtain the information of color point clouds in three-dimensional space.However,how to unify the color point cloud collected from multiple viewpoints in the same coordinate system is still a challenge.Therefore,an image descriptor based on deep learning is proposed,which is applied to 3D color point cloud registration to obtain the initial pose of point cloud registration with high precision.Firstly,according to the one-to-one correspondence between point clouds and images,3D color point clouds can be projected into images.Secondly,the convolutional neural network is used to extract the local features of point cloud key points,and the combined descriptor is formed by combining the directional gradient histogram.Thirdly,the matching point pairs of point clouds are calculated according to the calculated combined descriptor,and the transformation relationship between point clouds is obtained to realize the point cloud coarse registration.The validity of the proposed method is verified by comparing the actual 3D color point cloud data with various registration algorithms.

关 键 词：彩色点云数据 刚性配准 局部描述子 

分 类 号：Q-34[生物学] TP301.6[自动化与计算机技术—计算机系统结构]